FAB symposium

Barry Quinn PhD CStat
Director of Finance and AI research Lab
b.quinn@qub.ac.uk

FAB

At the Finance and AI Research Lab, Queen’s Business School, we recognize the application of artificial intelligence in financial services as a wicked problem:


A problems with combined degrees of conflict (Adelson et al. 2023), complexity (Kelly, Malamud, and Zhou 2022) and uncertainty(Dotan and Ravid 1985).

Mission

  • Exploring AI Frontiers in Finance: Tackling complex, ever-evolving AI challenges in financial services.
  • Innovative and Responsible AI Solutions: Advancing financial practices with robust, ethical AI applications.
  • Interdisciplinary Approach: Merging finance, technology, and ethics for comprehensive solutions.
  • Future-Focused Strategies: Prioritizing long-term sustainability and responsibility in AI integration.
  • Leadership in AI and Finance: Pioneering change, driving innovation, shaping an inclusive financial future.

Vision

  • Robust AI Model Development: Focusing on creating advanced and reliable AI solutions.
  • Cultivating Future-Constraining Frameworks: Building systems to address ‘super wicked problems’ as outlined by Levin et al. (2012).
  • Proactive Long-term Engagement: Actively considering the far-reaching impacts of our AI innovations.
  • Sustainability Commitment: Ensuring our AI advancements are sustainable over the long term.
  • Equitable Solutions: Striving for fairness and inclusivity in all AI applications.
  • Transparency in Innovation: Maintaining openness and clarity in our AI development processes.

Approach

  • Interdisciplinary Expertise: Merging finance, technology, data science, and ethics for holistic problem-solving.
  • Creative Problem-Solving Environment: Encouraging innovative thinking in tackling complex challenges.
  • Addressing Wicked Problems: Confronting intricate issues with a focus on innovation and forward-thinking solutions.
  • Shaping a Responsible Financial Ecosystem: Moving beyond efficiency and profit to foster inclusivity and responsibility in AI-driven financial services.
  • Resonating with Modern Financial Services: Tailoring solutions to the nuanced needs of a rapidly evolving, AI-integrated financial landscape.

Activities

  • Financial tail risk, advanced analytics and artificial intelligence Knowledge Transfer Project (UKRI) with Funds-Axis Ltd.

  • Leveraging Artificial Intelligence to enhance and understand regulatory compliance in the investment management industry with Funds-Axis Ltd.

  • Towards a trustworthy banking approach to AI implementation: Balancing membership trust with operational performance with Credit Union Development Association.

Market manipulation in capital markets

  • A collobarative research project with
  • Weilong Lui (Shenzhen University)
  • Fearghal Kearney (FAB at QBS)
  • Jesus Martinez Del Rincon ( GII at EEECS)

Global capital markets

  • Figure 1 is high-level view of financial markets’ importance.
  • Trust and integrity are foundation to their functioning of these markets.

Figure 1: Relative size of capital markets (2022)

30 days in US equity trading

Data sourced from BATS

Economics 101

Figure 2: Economic view of markets

Paradox of the markets

Paradox of the markets

Paradox of the markets

  • Much like Groucho Mark’s refusal to join a club that would have him as a member,

  • Everyone should refuse to transact with anyone willing to transact with them!

Financial market participants

Speculators

Economics 101 market microstructure theory

  • Now, the large number of perfect competition ideals breaks down, and market participants’ behaviour becomes strategic:
  • This idea forms the basis of the theory of market microstructure.
  • Taking into account that their actions can now change the price (can endogenously determine prices)
  • We have seen how the largest part of trading costs is the adverse effect on prices; called the implementation shortfall.
  • The theory allows us to classify a illegal trader as an insider (some with private information).
  • Most of the time, there is no one acting as Auctioneer.
  • Most trading occurs during continuous trading sessions.
  • In these interactions, the market procedures and rules matter very much.

Fraud and its evolution

  • Fraud, defined as criminal deception for unjust advantage, has evolved with technology (Bolton and Hand 2002).

  • Fraud detection usually works along side fraud prevention, where there is a necessity for detection methods when prevention fails.

  • Classic prevention methods include:

  • Elaborate designs on banknotes such as fluorescent fibers, multitone drawings, watermarks, laminated metal strips, and holograms.
  • Personal identification numbers (PINs) for bankcards.
  • Internet security systems for credit card transactions.
  • Subscriber Identity Module (SIM) cards for mobile phones.
  • Passwords for computer systems and telephone bank accounts.
  • Fraud detection is a continuous and evolving process, necessary because criminals adapt and develop new strategies to circumvent existing detection methods.

Market manipulation definitions

  • Spoofing and closing price manipulation are both forms of market manipulation but they differ in their methods and objectives.

  • Spoofing is an especially prevasive problem in US stock markets where 97% of orders are cancelled before they trade (Khomyn and Putniņš 2021)

  • JP Morgan paid over $900Million in fines for spoofing activity in the commodities markets during 2008-2016 (Debie et al. 2023)

What is Spoofing

  • Definition: Spoofing involves placing large orders to buy or sell a security with no intention of executing them. The goal is to create a false impression of high demand or supply, thereby manipulating the market price.
  • Method: Traders place large orders that they plan to cancel before execution. These orders are usually placed just outside the current bid or ask price to influence other market participants without actually executing any trades.
  • Objective: The primary goal is to influence the behavior of other traders. For example, by creating an illusion of increased demand, the spoofer may drive up prices, at which point they can sell at an artificially inflated price.

Price impact of spoofing

- This is an actual FINRA manipulation case from 2016(Zhai, Cao, and Ding 2018)

  • In this example quote stuffing was used to create an impression of strong buying interest.
  • Market status: Ask price: 101.35; Bid price: 101.24.
  • The purpose of the manipulator is to execute a sell order at a high price of 101.32.
  • Step 1: Manipulator add a bona fide sell order with 1000 shares at 101.32.
  • Step 2: The manipulator adds a series of non-bona fide buy orders to push up the bid - price to 101.31.
  • Step 3: Some investors, encouraged by the (fake) bid price changes, responded to the sell order. Thus, some of the 1000 shares of the sell order are executed at the expected price of 101.32.
  • Step 4: The Manipulator cancelled all the buy orders. Thus, the bid price went down to 101.24.

Data

  • We have extracted information from the official announcement of administrative penalties.

  • Which stock was manipulated on which day?

Important

  1. Declaration and transaction of shares of “Yunnan Copper” on 25 October 2010. On 25 October 2010, the “Su Yan Xiang” account group declared 15 buy orders for 6,350,000 shares, accounting for 5.42% of the total number of buy orders in the market throughout the day, ranking first in buy orders. 9.25%.

Data

  • Timeframe of the administrative penalty decision letter:From 2014-01-01 to 2022-12-31 (165 files)
  • Timeframe of manipulation cases: From 2012-01-01 to 2020-12-31. (287 stocks)
  • If a stock was manipulated on a particular day, we downloaded a full year’s worth of data for that stock.
  • Each day of data for a stock is considered a sample.
  • The manipulated date is marked as abnormal data.

Sample

  • For each sample, a 5-minute interval of high-frequency trading data is taken as input data.
  • A Big dataset of labelled manipulations.

Sample

  • We obtained 43,045 samples, of which 1872 are manipulation cases.

  • Input: One day time series data for one stock.

  • Matrix (number of time steps x number of features)

  • 5 minutes sampling, approximately 50 time steps per day.

  • Output: Normal (0) or abnormal (1).

Model

  • We use a Long Short Term, a non-stationary algorithmic approach to financial time series data.
  • One of the most popular off-the-shelf models to predict financial markets (Josh et al. 2022)

## Results

## Results

## Next steps

  • Sample imbalance
  • Develop a trade BERT framework for market manipulatio in spoofing
  • Using market mircostructure theory to inform financial feature engineering

Concluding remarks

  • Trader’s who act illegal can be thought of as be informed in theory, which can help make statistical detection more tractable.
  • There is a huge statistical imbalance in any labelled datasets used to build models as market abuse cases are rare (less than 1% in most dataset).
  • Regulators must balance between public reporting of fraud cases and regulatory arbitrage of illegal traders.
  • More work is need to better understand the pyschology of a rogue trader.

References

Adelson, Caroline, Charlotte Kuller, Cate Tompkins, Ellora Sarkar, Samantha Price, and Marco Iansiti. 2023. “How Wicked Problems Drive Business Performance: A Review of the Academic Literature.” Harvard Business School Working Paper Series 23-064. https://www.hbs.edu/ris/Publication%20Files/23-064_f10377b2-89e6-4752-ac42-456c0f42136f.pdf.
Bolton, Richard J., and David J. Hand. 2002. “Statistical Fraud Detection: A Review.” Statistical Science 17 (3): 235–55. https://doi.org/10.1214/ss/1042727940.
Debie, Philippe, Cornelis Gardebroek, Stephan Hageboeck, Paul Leeuwen, Lorenzo Moneta, Axel Naumann, Joost M. E. Pennings, Andres A. Trujillo‐Barrera, and Marjolein E. Verhulst. 2023. “Unravelling the JPMorgan spoofing case using particle physics visualization methods.” European Financial Management 29 (1): 288–326. https://doi.org/10.1111/eufm.12353.
Dotan, Amihud, and S Abraham Ravid. 1985. “On the Interaction of Real and Financial Decisions of the Firm Under Uncertainty.” The Journal of Finance 40 (2): 501–17.
Kelly, Bryan T, Semyon Malamud, and Kangying Zhou. 2022. “The Virtue of Complexity Everywhere.” Swiss Finance Institue Research Paper No.22-57. https://doi.org/10.2139/ssrn.4166368.
Khomyn, Marta, and Tālis J. Putniņš. 2021. “Algos gone wild: What drives the extreme order cancellation rates in modern markets?” Journal of Banking & Finance 129 (August): 106170. https://doi.org/10.1016/j.jbankfin.2021.106170.
Zhai, Jie, Yong Cao, and Xiaohui Ding. 2018. “Data Analytic Approach for Manipulation Detection in Stock Market.” Review of Quantitative Finance and Accounting 50: 897–932. https://doi.org/10.1007/s11156-017-0650-0.